Exploring the Rise of Mortgage BorrowingAmong Older Americans

J. Michael Collins, Erik Hembre, and Carly Urban∗

January 23, 2020

Abstract

Between 1980 and 2015, older households in the US more than tripled the use ofhome mortgage debt. Rather than using owned homes as a source of imputed rent,older households are borrowing against home equity, with loan terms that exceed theirexpected life spans. Using several data sources, we explore the rising use of mort-gages among elderly homeowners. Rising mortgage borrowing provides low-wealtholder households with increased liquid assets, but it does not appear to be mean-ingfully associated with increases in loan defaults. This trend of elderly mortgageborrowing is not explained by increasing levels of income or cohort demographicshifts, but is linked to a rise on ownership of homes by older households in general.However, changes in subsidies associated with mortgage debt partially contribute todifferential increases in mortgage use by older households.

JEL Classification Codes: D12, D14, D15Keywords: Household Economics, Consumer Debt, Personal Finance, Personal Savings, Retire-ment

∗Hembre gratefully acknowledges financial support from the Steven H. Sandell award from the BostonCollege Center for Retirement Research. The authors thank seminar participants at DePaul University,the Household Finance Seminar Series at the University of Wisconsin-Madison, and the Urban EconomicsAssociation.

1 IntroductionIn the United States, the share of households headed by someone age 65 or older is in-creasing to the point where one in five household heads will soon be labeled as ‘elderly’(Poterba, 2014). At the same time, older households are holding an increasing level of debt(Consumer Financial Protection Bureau, 2014; Vornovytskyy et al., 2011). According toUS Census data, 3.6 million more households aged 65 and older had a mortgage in 2015than in 2000, an increase of 39 percent. Meanwhile, households headed by someone under40 had four million fewer mortgages over the same time period. Part of the increase is dueto a rise in homeownership rates among older households. However, the rate of mortgageusage has nearly tripled, from 13 to 36 percent, among elderly homeowners since 1980.

A traditional life-cycle savings model predicts that individuals borrow at younger ages,then pay off debt and decumulate assets in retirement. Mortgage debt used to buy a homeis an example—households borrow at younger ages to buy a home, pay off the loan andthen use that asset to consume housing as they age, providing imputed rent. However,the rate of older households holding a mortgage is rising relative to previous cohorts. Forexample, Census data in Panel A of Figure 1 show that the rate of holding a mortgagenearly quadrupled from 10 percent of households near age 70 in 1980 to almost 40 percentby 2015. Home mortgage loans are typically the largest loans that households take on, andpaying off these loans creates home equity and frees up cash flow for other consumption.However, the rate of seniors holding on to mortgages as they age is rising. This begs thequestion, why are older households increasingly holding mortgage loans?

Some media reports argue that rising levels of debt for older Americans presents aserious problem (Pham, 2011). A report by the Consumer Financial Protection Bureau(2014) claims that “rising mortgage debt is threatening the retirement security of millionsof older Americans.” Recent work by Mayer (2017) highlights increasing debt amongretirees relative to financial assets, reporting that in 2012, 40 percent of homeowners aged65 to 69 have more mortgage debt than they have financial assets, up from 28 percentin 1992. If older homeowners hold mortgages and are burdened with making mortgagepayments as they exit the workforce in retirement,1 they have an added monthly expense,potentially crowding out consumption of other welfare enhancing goods, such as healthcare, food, and prescription drugs.

However, given longer life expectancy, improved health status, and extended laborforce participation rates of older workers, households may be optimally choosing to main-

1We acknowledge that retirement from work and elderly status are both defined continuously, withhouseholds engaging in variable levels of work even at the older ages. Our assumption is that households65 and older are more likely to be retired from work, and more likely to receive pension and Social Securityincome than market income.

1

tain mortgage debt later in life. The housing boom of the 2000s may have encouragedborrowers to substitute away from higher cost debt, such as auto or credit card debt, tolower cost mortgage debt (Brown et al., 2015). It is also plausible that each new cohort ofelderly households is being shaped by generational shifts. As the number of older peoplewho own homes increases, and the composition of the elderly home owning populationchanges, the probability of older people holding a mortgage may also shift. The propen-sity for the elderly to hold mortgages may simply be a function of differences in eachcohort’s wealth and income trajectory, for example. Moreover, income tax, estate tax andother incentives may favor holding mortgage debt, combining with life-cycle factors andbequest motives, to make mortgage holding more favorable for more recent elderly cohortsthan prior generations.

As household heads age, they reconsider the costs and benefits of owning versus rent-ing a home. Older households are less location constrained, since they are less likely tobe commuting to work and are unlikely to have school-age children. Given the lower op-portunity costs of time for people out of the workforce, these households also have lowertransaction costs of moving.2 Combined with greater flexibility on housing location, thesefactors could encourage households to sell their homes and then rent or downsize to alower cost property. Selling a home can free up home equity to be used for investmentor consumption. On the other hand, the income of older households is less likely to risewhen rents rise. By owning a home, households hedge against the increases in housingcosts that renters face. For wealthier older households, owning housing provides an assetthat may complement other assets in their portfolio. These financial factors might combineto encourage home owning in old age.

Even when older households decide to become or remain homeowners, they can de-cide to own the home outright with no loan, or have mortgage debt against their home.Mortgage usage among younger households is often rationalized as a way to smooth con-sumption during earlier working years. Older populations have likely already aged out ofpeak earning years, making the use of a loan less compelling. At the same time, olderhouseholds are less likely to experience unemployment spells and tend to have relativelystable income from Social Security and any annuities or defined pension income. Thismight result in older households perceiving the payment default risk of a mortgage asbeing lower than for younger households with more income volatility.

For some older households who own homes, having a mortgage may actually be drivenby financial need. The boom-and-bust housing cycle of the 2000s had a dramatic impacton home values and access to mortgage debt overall. As shown by Bhutta and Keys (2016),unprecedented real housing gains in the boom period, combined with low interest rates, ledto high rates of equity extraction among homeowners in general. As home values plateaued

2Dietz and Haurin (2003) describe home sale transaction costs.

2

and then declined, homeowners in many parts of the country faced large losses and evenhome foreclosure. While older households were less likely than average to experience aforeclosure, the price decline still exposed their household balance sheets to the full bruntof the housing bust. Lower home values may have also delayed older households’ plans tosell their homes to pay off their mortgage. Older households may have also been indirectlyimpacted by the recession if their adult-aged children needed in-kind or cash inter vivostransfers, which could have substituted wealth otherwise that would have been used topay off a mortgage. Reverse mortgages, which are specifically designed to facilitate olderhouseholds to tap into home equity to fund consumption, have not been in high demand(Shan, 2011). Instead older households may simply be extending existing mortgage debt.

This paper provides a deeper understanding of the prevalence of mortgages amongolder households over time. In addition to showing relative changes in the use of mort-gages by subgroups within the elderly population, we explore tax policy and other factorsthat may have pushed higher relative rates of mortgage borrowing among older house-holds.

This study complements several recent studies. Mayer (2017) documents evolvingtrends in homeownership and mortgage finance among older adults over the past severaldecades. He reports that while homeownership rates among older adults have been increas-ing over time, mortgage usage has also been increasing without a corresponding increasein financial assets. The author suggests that this is potentially a signal that elderly house-holds experience less financial stability as they take on more debt. Lusardi et al. (2017) usedata from the Health and Retirement Study from 1992, 2004, and 2010 to show that recentcohorts entering into retirement have taken on more debt than in the past, largely to livein higher-priced homes with a lower equity percentage (also know as higher loan-to-valueratios). The authors did not follow cohorts past the initial years of exiting the labor force,however. Brown et al. (2016) explore the prevalence of mortgage holding by age over timeusing data from the Federal Reserve Bank of New York/Equifax Consumer Credit Panel(CCP). Using these administrative data, the authors also show rising debt levels amongolder households. These papers all raise the same basic questions we pose, and find thatolder homeowners are more likely to carry a mortgage and take on more mortgage debt inrecent years than they did in prior decades.

We build upon these studies, exploring the heterogeneity in the use of mortgages byage cohorts over time, including bequest motives, education level, pensions and wealth.We then explore potential rational explanations for the increased debt load by these house-holds, including state tax incentives, regional unemployment rates, and relative rents.Building on prior studies, we estimate the likelihood of holding a mortgage by comparingolder households to slightly younger cohorts.

We find that relative to 50 to 64 year olds, households with heads aged 65 to 79 in-

3

creased their mortgage usage by ten percentage points between 2000 and 2016. Evenconditional on homeownership, mortgage borrowing has risen considerably among thisgroup. These older households also have higher home values and home equity than priorcohorts, as well as higher rates of mortgage refinancing. The rise in the relative use ofmortgages among older households is in part due to rising local unemployment rates, par-ticularly the rise in unemployment during the Great Recession. Local area rents relativehousing prices also are a factor in the choice to borrow.

We also consider the role that tax incentives, through the mortgage interest deduction,may have on increased mortgage borrowing among older households. We build on previ-ous literature examining how tax incentives shift consumption behavior, such as DeFuscoand Paciorek (2017), Dunsky and Follain (2000), Hilber and Turner (2014) and Ling andMcGill (1998). We use variation in an individual’s subsidy amount based on changes instate income tax rates, tax brackets, treatment of mortgage interest, and treatment of othertax deductions by state and year to estimate relative incentives for mortgage use by agecohort. We estimate about half of the increase in mortgage usage by older households isrelated to tax policy.

Mortgage borrowing could have economically significant implications for the financialwellbeing of retired households. Given historically low mortgage interest rates, house-holds may simply be exploiting financial arbitrage between investment and borrowingrates. Households may be able to increase long-run consumption by “borrowing low” and“investing high.” Indeed, Goodman and Mayer (2018) provide evidence that even overthe boom-bust housing cycle, owning a home can be a financially advantageous strategygiven house price patterns. However, holding a mortgage is risky. The house and propertyprovide the collateral for the loan. If borrowers cannot maintain payments and ultimatelydefault on their loans, they risk losing a significant financial asset as well as their source ofhousing. If households are borrowing primarily to tap home equity for consumption ratherthan investment, and they fail to keep up with debt payments, more households may be atrisk of experiencing financial problems in older ages than they did in prior generations.

We do not find evidence of older households being at greater financial risk from de-faults as they age and continue to hold mortgages, relative to other age cohorts or priorgenerations. Still the general pattern remains important to monitor, especially as marketcycles, interest rates and home prices fluctuate. It may simply be a matter of older house-holds’ shifting preferences and risk tolerances, combined with credit markets willing toextend a supply of mortgage loans to elderly consumers.3 A shock to retirement incomes,health care or other consumption costs, or a prolonged credit contraction could make theseleveraged households more vulnerable to hardships.

3Fair Lending regulations in the US also restrict lenders from denying credit solely on the basis of a loanapplicant’s age.

4

This paper continues by providing a brief background on prior studies on homeown-ership and mortgage use among the elderly, mainly in the US context. We next provideinformation on the data sources and estimation strategies, followed by a series of visualand tabular representations of these estimates. We systematically show the relative use ofmortgages for older households relative to slightly younger cohorts, by sub-population orgroup characteristics, in order to narrow in on heterogeneity in mortgage use. We thenconclude with a brief discussion of the results and implications for the field.

2 BackgroundYounger households with higher expected future income and little current savings canuse debt to smooth consumption over the life course. As households’ ages and incomespeak, these households can pay down and eliminate debt while saving for retirement. Theprediction from standard models is that older households will hold relatively less debt thanyounger households do.

As households age, they can select one of three housing options: (1) Own a homeoutright, without a mortgage; (2) own a home with a mortgage; and (3) rent a home, eitherfor cash, or a no-cash rent, such as living with relatives. An increasing number of olderhouseholds are choosing option (2), which motivates our analysis.

Mortgage debt is used to finance property, typically an owner-occupied residence thatoffers a stream of housing consumption to the owner.4 The standard US mortgage loanhas a 30-year amortized repayment term. This means that a mortgage originated when ahousehold is 30 would naturally terminate by age 60, assuming no refinancing, prepay-ments, or selling of the home. Households who take out a mortgage at older ages couldstill pay off the loan ahead of schedule, use savings to pay off the balance, or sell the hometo pay off the loan.

Older homeowners who lack assets to pay off their mortgage can sell their homes, andthen use the proceeds to finance consumption while renting a home. Renting may also beattractive as homeownership exposes a large portion of household wealth to the occasion-ally volatile local housing markets. Housing-rich but income-poor older homeowners mayalso be burdened by property taxes (Shan, 2010) and could downsize their living space byselling their home and renting a smaller housing unit. Rental properties typically requireless physical maintenance, which could be attractive to older households with physicallimitations (Golant, 2008a,b).

4Housing consumption value represents the flow utility of living in a home and is often approximated atthe imputed rental value of the home.

5

Despite these potential benefits, the general trend is not for older households to tran-sition from homeownership to renting. Painter and Lee (2009) study the housing tenuredecisions of older households using the Panel Survey of Income Dynamics and concludethat age does not directly relate to households transitioning from ownership to renting.Shocks such as having a lower health status or becoming a single head of the householdare more predictive of housing transitions than age alone. While there have been changesin the delivery of health care that may facilitate older households staying in their homesin spite of health shocks, there is also not strong evidence health care policies are con-tributing to more homeownership among the elderly (Engelhardt and Greenhalgh-Stanley,2010).

Older households may prefer homeownership since it provides more predictable hous-ing costs than rental markets, which are subject to annual contracts (Sabia, 2008; Sinai andSouleles, 2005). Aging households may also prefer owner occupied homes to be able tocontrol the level of investment they make in home maintenance. Aging homeowners cansmooth consumption by forgoing home maintenance, essentially reducing the equity valueof the home over time and consuming what would have otherwise been spent on home re-pairs (Gyourko and Tracy, 2006). Another factor that may contribute to homeownershipof older households is the steady rise in Social Security benefits, which reduces incomeuncertainty in retirement (Engelhardt, 2008).

Of course, the composition of the elderly population who own homes is also changingover time, especially for the Baby Boomers cohort which has greater wealth and betterhealth than prior age cohorts. This could also affect their probability of holding a mortgagewhen a homeowner. A number of studies (Brown et al., 2016; Copeland, 2015; Lusardiet al., 2018; Vornovytskyy et al., 2011) have shown absolute and relative increases in debtlevels by age. The growth in mortgage debt exceeds what might be explained by risinghomeownership rates in a population with longer life expectancy alone. The same trendis occurring in other countries, including among retirees in Canada (Bedard et al., 2018).Economic theory and descriptive studies suggests that older households will tend to remainin owner occupied homes as they age (Golant, 2008a,b). The question remains, however,why are older homeowners more likely to use a mortgage in recent years?

There are several reasons why home-owning households may strategically maintainmortgage debt as they age. First, financial planners counsel households on the tradeoffsof paying off a mortgage as part of retirement planning (Nason, 2017). Since mortgageinterest rates are often a low cost form of borrowing, households can leverage their port-folios by arbitraging the difference between the costs of borrowing and real rates of returnwhen investing in markets. Such strategies can be risky, since the home is collateral for theloan and the asset values of investments and housing can decline. It is unclear whether theaverage older household would have the ability to effectively borrow to invest, especially

6

given low levels of measured financial literacy (Lusardi et al., 2017). Taking on debt mayalso expose a couple to the risk of one partner dying, leaving the surviving partner withmortgage debt but lower pension and Social Security income. While households couldhedge this risk with insurance, only about one in three elderly households had cash valuelife insurance in 2016 (Bricker et al., 2017).

A second reason that older households may maintain mortgage debt as they age isbecause of tax incentives. At the federal level, and in most states, mortgage interest isdeductible from income taxes. Several studies, including Hanson (2012) and Hilber andTurner (2014), examine the influence of the mortgage interest deduction on housing deci-sions. There is little evidence that the mortgage interest deduction has an impact on theextensive margin of the homeownership rate, although Hanson (2012) finds an intensivemargin effect of the mortgage interest deduction on home size. While there are other taxincentives related to housing, such as the ability to deduct property taxes from income,the exclusion of capital gains from income taxes, and imputed rent from taxation, these allencourage homeownership, though not specifically borrowing or extending a mortgage.

Mortgage subsidies should prioritize mortgage debt over shorter-term consumer debt,such as automobile loans and credit cards, which typically charge higher rates. The sub-sidies lower the effective, after-tax interest rates of mortgages, making it less costly tofinance current consumption through mortgage debt than to draw down financial invest-ments. Amromin et al. (2007) show that the tax-exemption of certain retirement savingsaccounts means households might be better off if they continued to borrow using a mort-gage loan while putting money into retirement accounts. The authors do not find evidenceof this behavior, however. Part of the issue is that the mortgage interest tax deductionrequires itemizing, which typically only applies to higher-income borrowers (Poterba andSinai, 2008).

A third reason that households may hold mortgages as they age is the need to borrowto fund consumption. Poterba et al. (2011) show a pattern that is consistent across studies:many households have little savings even as they approach retirement. Home equity is oneof the primary stores of non-pension wealth, especially for low-income and less wealthyfamilies (Bricker et al., 2012). Home equity may be one of the few ways for householdswith little other savings to smooth spending or respond to financial shocks. Among allhouseholds age 65 to 70 in 2008, real estate represented 48 percent of non-Social Securityor pensions wealth (Poterba et al., 2011). Several studies show that households often usehome equity as a source of liquidity. For instance, Bhutta and Keys (2016) use a panelof consumer credit reports to document high rates of home equity extraction as a result ofincreasing house prices and low interest rates in the early 2000s. They further showed thathouseholds use this equity to pay down other consumer debts. Davidoff (2010) shows thathome equity may substitute for long-term care insurance, and may therefore be a useful

7

financial planning strategy. However, the value of home equity can be volatile, whichpresents a substantial risk for homeowners. Lusardi and Mitchell (2007) even suggestedthat some households in retirement have less home equity due to failure to plan ahead topay off their mortgages.

Moulton et al. (2017) studied home equity conversion mortgages, or reverse mort-gages, which is a type of loan designed for liquidity constrained borrowers age 62 or olderwhere the balance is not due until the home is sold or the borrower is deceased. Theycompare otherwise similar reverse mortgage borrowers to borrowers of traditional (for-ward) home equity mortgages, finding that reverse mortgage borrowers are more likely topay down existing debt. While reverse mortgages are subsidized by the US Department ofHousing and Urban Development, only a small fraction of senior homeowners utilize theproduct (Moulton et al., 2017; Shan, 2011). In part this is due to the lowest income house-holds not having sufficient home equity to produce significant financial benefits (Venti andWise, 1991).

Overall, the literature suggests that homeownership among the elderly is likely to bepersistent across successive cohorts of the same age, but offers few insights into why morerecent cohorts of older homeowners are utilizing mortgages at higher rates than in thepast. Our study examines mortgage holding by older homeowners over time compared toa slightly younger age cohort.

The observation of rising use of mortgages among the elderly is important if this risinguse of debt may result in negative consequences for the well-being of elderly households.For example, Brown et al. (2019) examine factors for hardships (poverty, use of meantested benefits, food insecurity, or a drop in wealth), including a simulation of the effectsof rising debt. They find that debt is not a strong factor related to hardships, which theauthors suggest is because people who take on more debt at earlier ages tend to be moreeconomically secure than those who take on less debt.

Selection into mortgage debt could be positive, reducing the risks of hardships. Butricaand Karamcheva (2018) find that higher debt levels delay claiming of Social Security, asimilar finding to a working paper by Moulton and colleagues (Moulton et al., 2019). Thisis also consistent with positive selection into mortgages and financial stability (at leastmeasured by labor force participation or the ability to delay Social Security claiming dueto having other assets).

Zhao and Burge (2017) find that housing wealth may encourage people to exit thelabor force sooner as higher debt levels reduce net housing wealth and therefore delayretirement. At some level mortgage debt may not be sustainable. Bian (2015) showed thataging households are more likely to downsize or sell their home if their mortgage debt islarger relative to home value, at least at higher levels of the debt-to-asset ratio.

Brule and Ravazzini (2019) studied subjective well-being (self-reported) of elderly

8

people in Europe. They found that people report higher subjective well-being when theyhave more liquid assets than when they have wealth locked up in assets like housing.They also found mortgage debt levels have little relationship to subjective well-being.Other studies take a more pessimistic view of debt among older households (Jappelli etal., 2013), including links between debt levels and health (Argys et al., 2016). Still otherstudies link debt levels to particular forms of anxiety and self-reported debt-related stress(Drentea and Reynolds, 2015; Dunn and Mirzaie, 2016).

It is notable across the literature that debt, especially mortgage debt, is treated as beinga useful strategy to achieve higher financial well-being in some studies, or a risky threat tofinancial stability in other studies. The use of mortgage debt, relative to other borrowingoptions, clearly has advantages in terms of the cost and offering a long-term repaymentstructure. As described by Zinman (2015), researchers need to be cautious in how toconsider consumer’s use of debt. For some households, too little debt can limit valuableinvestments; for others too much debt can result in default. The optimal level of debt byage may be determined by a range of factors. To understand the use of debt, we need tobetter understand the heterogeneity of circumstances that elderly households face. Withthis in mind, we turn to several widely-used datasets to document the use of mortgage debtby older households based on observed characteristics.

3 DataThis paper aims to describe trends in mortgage holding across age cohorts using the Surveyof Consumer Finances (SCF) and the Health and Retirement Study (HRS), as well asgeneral population patterns in Census data.

The SCF data provide a comprehensive look at households’ personal finances. Thestudy is repeated cross-sectional data administered by the Federal Reserve Board in con-junction with the Department of the Treasury with triennial surveys since 1983. The great-est advantage of the SCF relative to the HRS and Census data is the detailed collection ofhousehold financial attributes including total assets and net worth, mortgage usage, homeequity, pensions, savings, and delinquent debt payments. While the SCF contains ques-tions not included in the HRS or the Census data, it does not provide geographic identifiers.Since the earliest SCF surveys do not contain our variables of interest, we begin our anal-ysis with the 1989 survey for most estimates. We provide estimates for all households andthen only for households who own homes.

The HRS is a biennial panel survey designed to track the demographic characteristics,health status, and financial assets of households 50 years and older in the United States.Beginning in 1992 with an initial cohort of 12,652 respondents, the HRS’s primary re-

9

spondents were born between 1931 and 1941. To cover the full age distribution above50 years, the HRS has added additional cohorts over time. Of particular interest for thisstudy, the HRS collects information on housing tenure status, mortgage status, and pay-ment amount. We utilize restricted-access HRS data, which include geographic locationinformation, allowing us to include housing market variables such as tax environments,rent-to-house price ratios, and unemployment rates. In our analysis, we restrict our sam-ple to primary respondents between the ages of 50 and 100, between 1994 and 2014. Weprovide estimates for most specifications for all households, and then only households whoown homes.

We also use tabulations of the 1980, 1990, and 2000 PUMS decennial Census, and2005, 2010, and 2015 American Community Survey to document the rate of older house-holds owning homes and having a mortgage. We use a random sample of about a half-million households from each wave of these Census data, again excluding householdswith heads younger than age 50, and older than age 100. There is a total of six waves, cov-ering a 35 year period. This dataset gives us the longest time horizon with which to studymortgaged rates. The Census data include housing tenure (rent, own with a mortgage,or own without a mortgage). We observe the age of household head, household income,household size, marital status, and state of residence. However, the Census data do notinclude information regarding household assets.

3.1 Trends in Mortgages, Homeownership, Wealth, and RetirementMortgage debt holding among the elderly has risen substantially over the past thirty years.Figure 1 illustrates this trend in the rate of holding of a mortgage and homeownership ratesfor all households using data from the Census. Panel A shows that households decreasemortgage usage as they age past their 40s and 50s. However, households older than 65 arecurrently much more likely to have a mortgage than in previous decades. For instance, in1980 only 4 percent of households aged 80 to 84 held a mortgage. By 2015, that rate hadquadrupled to 17 percent.

Figure 1, Panel B, suggests that increased mortgage borrowing is partly due to increas-ing homeownership. Between 1980 and 2015, homeownership rates actually decreased forthose under 50 years old, likely due to family formation and career trends for this group.5

While homeownership rates steadily increase as households age, they peak around 80 per-cent when households are in their 60s. The interesting pattern occurs after the intersectionof homeownership rates in 2015 (and to a lesser extent, 2000) with rates of previous yearsnear age 65. Younger households in 2015 are less likely to be homeowners than in previ-ous years, and older households are more likely to be homeowners than in previous years.

5 Gruber et al. (2017) point out a comparable trend in Denmark.

10

In fact, the largest absolute change in homeownership rates between 1980 and 2015 is forthe 80 to 84 age group, which has increased homeownership rates by 18 percentage pointsfrom 64 percent to 82 percent.

An immediate question arises: are these trends indicative of a movement towards aneed to retire at older ages? Figure 2 uses the HRS waves from 1998 to 2014 by ageto illustrate trends in wealth (in constant 2014 dollars) and retirement rates again for allhouseholds—not just homeowners. The general pattern in Panel A is that wealth accu-mulates as household heads approach retirement ages, then plateaus or declines. Wealthis just over $100,000 lower on average in the 2010 to 2014 period than it was during the2004 to 2008 period, reflecting the economic cycle. Only the 1998 to 2002 data show adecline in wealth with age. There are two other notable trends. First, wealth accumulationoccurs much later in life for the 2010 to 2014 cohort than it did for the 1998 to 2002 and2004 to 2008 cohorts. Second, for the earlier cohorts, wealth decreases more rapidly fromage 70 to age 90 than it does for the most recent cohort. Figure 3 shows the same patternsby wealth quintile. Wealth accumulation levels at the lowest quintile are low in general,and lower and happening much later in life for the 2010 to 2014 period. Wealth persistsinto later ages at similar rates across time periods at the median and highest quintiles. Thepattern of the lowest wealth group could be consistent with a need to borrow more at laterages.

Figure 2 Panel B displays the share of households who report being retired from work.Here the patterns are very similar across age cohorts where most households transition toretirement between ages 60 and 70. Changes in wealth or retirement do not seem to havechanged in recent years by age cohort the way that wealth holding has. These generalaverages could still be masking other characteristics of today’s seniors that could explainthese trends. We next turn to a more formal analysis of differences by cohorts over time.

4 Empirical StrategyTo determine the change in mortgage prevalence across cohorts and over time, we estimateEquation 1 using three different datasets. Each data sample allows for different measures,including the geographic location of the respondent. We focus on household heads in twogroups, age 65 to 79 and, compared to those aged 50 to 64. We then plot the estimateddifference by data year. The estimated difference between the 80 to 100 year olds and the50 to 64 year olds are shown in the Appendix. We estimate α1 in Equation 1 for eachsample year j, where j represents each year available in the three surveys, to begin tounderstand the decrease in the mortgaged rate gap between older and slightly less olderhouseholds in recent decades.

11

Yim = α0 +α1A2i +α2A3i +β1Xi +(δm+)εim (1)

In Equation 1, Yim is a binary variable equal to one if individual household i currentlyhas a mortgage and equal to zero otherwise. We choose a linear probability model forease of interpretation, but our results are comparable if we compare marginal effects froma logit. Xi includes number of children, logged total income, a high school educationor less dummy, race indicators (white or non-white), net worth quartiles, an employmentindicator, a marital status indicator (married, single male, and single female), and net worthquantiles by age cohort by year. In the HRS sample, Xi additionally includes bequestmotives are proxied by the self-reported likelihood of bequeathing at least $10,000 anddummies for the self-reported health status on a 1 to 5 scale.6

A2 and A3 are the second and third age cohorts comprising those ages 65 to 79 and80 to 100, respectively. The excluded age group is households with heads age 50 to 64.When we use HRS and Census data, we include δs, MSA fixed effects.7

The coefficients represent the gap between the prevalence of mortgages for 65 to 79year olds and 50 to 64 year olds. If the estimates by year are less negative, then olderhouseholds are closing the gap and becoming more likely relative to younger householdsto have a mortgage.

5 FindingsWe start by estimating a baseline rate of mortgage borrowing using the specification inEquation 1 with no controls. We then add control variables Xi. Further, we explore addi-tional outcomes that changed over this period (homeownership, mortgages conditional onhomeownership, home values, refinance rates, home equity, and equity extraction). Next,we explore heterogeneity in the changing mortgaged rates of older households. Finally,we estimate the effect of three economic or policy variables on the likelihood of holdinga mortgage over this period for each age group. As most of our analysis is descriptiveand intended to portray national averages we use sampling weights when computing esti-mates. When estimating the causal effect of tax rates, rental markets, and labor markets onmortgage usage we do not use sample weights since we are not concerned with identifyingpartial effects or endogenous sample selection.8 While we present nearly all of our results

6The SCF does not contain questions on bequest motives or health.7Neither MSA nor state identifiers are available in the SCF.8See Solon et al. (2015) for a more thorough consideration of the appropriate usage of using sampling

weights.

12

graphically, we provide estimated coefficients in the Appendix. We show estimates formost specifications for all households, and then only households who own homes.

Each figure displays the estimated gap between mortgage holding of households withheads age 50 to 64 versus those age 65 to 79 and corresponding 95% confidence intervals,for each survey year. These specifications include no controls beyond the age dummies;the HRS specifications include MSA-level fixed effects. These figures tell a general storyof how the gap between older and slightly younger generations’ mortgaged rates has nar-rowed over time.

5.1 Adding Controls for Wealth, Health, and Bequest MotivesThe rise in holding mortgages among older households may be explained by changesin demographics, wealth, health, or preferences to bequest money to children. Figure4 explores this question, where we include controls for demographics, wealth (SCF andHRS), health (HRS), and bequest motives (HRS) in addition to the demographic controls.Panel A explores the SCF data, which controls for all variables in Xi in Equation 1. Thetrend is a closing gap such that 65 to 79 year olds are borrowing at similar rates to 50 to64 year olds by 2016, including controls. Panel B, shows the HRS data and additionallyadds controls for self-reported health, bequest motives, and wealth quantiles. While thetrend still shows that 65 to 79 year olds are less different from 50 to 64 year olds in theirmortgage rates than in years past, the trend is quite flat after adding in controls.9

One explanation for rising mortgaged rates could simply be that older households aremore likely to own a home than they were before. Further exploration into the SCF datawith the same controls suggests that the gap in homeownership between 65 to 79 and 50to 64 year olds has been increasing steadily since 1989 (Figure 5). This trend persists evenwith the inclusion of controls. Increased housing consumption may indicate a growingpreference among older households towards owner-occupied housing. This contributes togreater mortgage borrowing as households may not have enough financial assets to owntheir optimal home size outright. However, Figure 5 also shows that the mortgaged rategap among owners narrowed by 2016, similar to the overall full sample (marked with an‘x’). This evidence supports the fact that the increase in mortgaged rates is in part due toan increase in homeownership by older households, though it does not fully explain whyseniors with homes are holding onto mortgages longer in life. Even allowing for morepeople age 65 and older owning homes, relatively more are also using mortgages.

9Note that the HRS data span more frequent but a shorter span of years than the SCF (1998-2014 vs.1989-2016). Thus, the HRS trends match the 1998-2014 trends in the SCF more closely than the overalltrend.

13

To further explore potential mechanisms for the puzzling trend, we change our out-come of interest to variables capturing the value of the home, the amount of home equity,the prevalence of equity extraction, and the likelihood of refinancing one’s mortgage inFigure 6. In Panel A, we show that from 1992-2004 home values across households withheads age 65 to 79 and age 50 to 64 were not different from one another. From 2007 to2016, the value of homes for the older cohort became greater than the younger cohort.This is reflected in greater home equity in Panel B.

Though 65 to 79 year olds hold more equity in their home than 50 to 64 year olds inrecent years, the two groups have been equally likely to extract equity from their homefrom 2007 to 2016 in Panel C. This was not always the case. The older cohort was lesslikely to extract equity in prior years. Finally, the gap in the likelihood to refinance steadilydecreased and in 2016 the two age groups were equally likely to refinance in Panel D, withcontrols. Taken together, Figure 6 suggests that older homeowners are accessing wealth intheir homes by borrowing.10

5.2 Exploring Heterogeneity by Age ProfilesAre the changing mortgage trends prevalent only for certain groups? We next seek toexplore heterogeneity in the changing trends of mortgaged rates. Figure 7 reports theestimated α1 coefficients from Equation 1, along with the estimated standard errors barsaround each coefficient for 65 to 79 year olds relative to 50 to 64 year olds.11

Panel A in Figure 7 uses an HRS question item about bequest intentions—here wemight expect that those households who plan to donate their wealth to their heirs wouldprefer to pass on a home free and clear of a mortgage. There does not appear to be anytime trend, although households without a bequest motive borrow at older ages similar tothose at younger ages, and the pattern of those with bequest motives shows lower relativemortgaged rates. Figure 8 Panel A shows these estimates only among homeowners. Herethe trend is flatter and shows few differences, however. In part this is because there arefew homeowners without bequest intentions, resulting in wide confidence intervals.

Panel B of Figure 7 uses the SCF to show those households where the head or spousehave a defined pension compared to those without pensions. Prior studies suggest thatrising levels of annuitized income may encourage owning a home and perhaps thereforemore borrowing (Engelhardt, 2008). There is selection in play, however, since people with

10Although not shown, we rule out mortgages on second homes; the magnitude of the second home rategap between age groups is close to zero in magnitude across our sample period.

11We continue to control for Xi in Equation 1 but exclude the variable of interest for each specification.When we split the sample by wealth, we continue to control for quantiles of those variables, but it reducesto one dummy variable instead of three.

14

pensions likely had careers and access to benefits that are different in multiple dimensionsfrom those without pensions. The trend over time is generally for a shrinking mortgageuse gap for both groups, however. Figure 8 Panel B shows no differences by pension andgenerally narrowing borrowing rates by age.

Panel C of Figure 7 uses the HRS to show that increases in educational attainment overtime are not likely to explain higher borrowing rates. The gap in mortgaged rates acrossthe two age cohorts for those with high school degrees or those who did not completehigh school are similar. The trend for both groups shows a closing of the gap, with moremortgage holding for 65 to 79 year olds relative to 50 to 64 year olds. If we expectthat education is a proxy for the type of employment, that is not a main reason for thedecreasing gap in mortgaged rates for younger and older households. Figure 8 also showsfew differences and relatively more borrowing among older households.

Panel D of Figure 7 uses the SCF to split the sample by those above and below medianwealth.12 Those below median wealth are largely closing the gap between the two agecohorts. At the same time, those 65 to 79 year olds above median wealth remain less likelyto hold a mortgage than 50 to 64 year olds with above median wealth. Figure 8, Panel D,which only shows homeowners, shows a less clear pattern; conditional on homeownership,wealth does not explain relative borrowing rates.

5.3 Tax Policies, Rental Markets, and Labor MarketsNext, we examine variation in the economic and policy environment over our period ofinterest to understand the magnitude to which they impact the gap between older andyounger cohorts’ mortgage borrowing rates. We explore subsidies from the mortgageinterest deduction (MID), local unemployment rates, and rent to price ratios (RTP) inEquation 2.

We predict that both tax benefits and relatively higher rental prices provide incentivesfor older households to have a mortgage. Local unemployment rates have less predictableeffects. Higher local unemployment trends could increase the rate that older householdsuse mortgages if homeowners to hold on to mortgages to help smooth income or to help outfamily members struggling to find work. Unemployment rates could decrease mortgagerates if slack labor markets and liquidity needs drive older people to sell their homes.Since many older households are not actively in the labor market, the effect of changes inunemployment trends could be minimal.

Mortgage borrowers who itemize their deductions on tax returns may deduct mortgageinterest from federal (and typically state) income taxes through the MID. Costing $83 bil-

12For example, in 2014 median wealth for the 50 to 64 age group was $60,000 and for the 65 to 79 agegroup was $146,800.

15

lion in tax revenue in 2017, the MID is the second largest tax expenditure for individualsin the federal budget (Burman et al., 2008). Lower-income owners benefit less from theMID than more affluent buyers do, both because lower-income households are less likelyto have enough deductions to itemize their returns instead of taking the standard deductionand because of the progressivity of the US tax code. We investigate the degree to whichsubsidies from the largest tax incentive for owning a home, the MID, vary by age. Sincewe include the MID average marginal tax rate in our regressions, we expect the relation-ship with mortgaged rate to be negative, as a lower tax rate increases the subsidy rate formortgages.

There is a body of literature studying the effects of the MID on homeownership, prices,and household finance decisions, as well as a broader literature on the effects of tax policyand homeownership (Gale et al., 2007).13

Hilber and Turner (2014) examine variation in the MID subsidy rate using a house-hold fixed-effects model to identify the effect of the MID on homeownership and houseprices, finding that the MID only increases homeownership among higher-income house-holds in less regulated housing markets. Sommer and Sullivan (2018) build a generalequilibrium framework to examine the relationship between the MID, rents, house prices,and homeownership and find, counterintuitively, that eliminating the MID would actuallyincrease homeownership.14 While previous research investigates the effects of the MIDon homeownership, home prices, and household debt, prior studies have not focused onthe heterogeneity of the effects of MID by age.15

We test for the influence of the economic and policy environments over our sampleperiod by estimating Equation 2.

Yimst = α0 +α1A2it +α2A3it +α5Pist +α6A2it ×Pist +α7A3it ×Pist+ (2)

β1Xit +β2MT Rst +ηm + γt +[δi+]εimst

13Variations exist in other countries, showing weak or mixed effects depending on the local context. Forexample, Jappelli and Pistaferri (2007) in Italy finds no effects of MID on homeownership or mortgage debtholding in Italy. Gruber et al. (2017) find large changes in tax subsidies for owner-occupied housing weredominated by demographic and other factors. In the US, Ling and McGill (1998) and Dunsky and Follain(2000) find that when tax rates are cut, the relative tax price of mortgages increases, and mortgage debt falls.

14A justification for the MID is the premise that positive externalities exist such that communities arebetter off with homeowners. However, there is little empirical evidence of this in practice. One exception isthat Engelhardt et al. (2010) find that when randomizing homeownership via subsidized savings householdsshow increases in home maintenance expenditures. There is little evidence supporting homeownership as ameans to facilitate other social benefits, however.

15One paper by Poterba and Sinai (2008) estimates simulated benefits of the MID by age, concludinghouseholds over age 65, on average, derive less value from the MID since their taxable income is relativelylow.

16

In Equation 2, we control for the same characteristics (β1Xit) from Equation 1, as wellas MSA (ηm) fixed effects, year fixed effects (γt), and the top marginal income tax rate inthe state in that year. In some specifications, we further include household fixed effects(δi) to compare households to themselves as they age and as policy environments change.The household fixed effects strategy is used by Hilber and Turner (2014) with the PanelStudy of Income Dynamics. Household fixed effects is not our preferred specification,however, since households age into and out of each age category over time. It becomeschallenging to identify the differential effects the policy by age for cohorts. Further, weestimate overall patterns across age groups and not within households throughout the prioranalysis. Although we show estimates both with and without household fixed effects,we prefer the estimates without household fixed effects. The estimates without householdfixed effects document the overall differences across age cohorts, as opposed to differenceswithin individual households in response to changes in policies. As in the prior estimates,the excluded age group are households with heads aged 50 to 64 years of age.

When we test for the effect of the mortgage interest deduction (MID) on mortgage us-age by age group, we follow Hilber and Turner (2014) by assigning Pist as the state averagemarginal mortgage interest tax rate as measured by the NBER.16 The estimated effect ofthe MID on mortgage use stems from within-state variation in the tax rate, as well as fromcross-state changes due to within-household mobility from state to state. The within-statevariation is due to changes in state income tax rates, tax brackets, treatment of mortgageinterest, and treatment of other tax deductions. To alleviate concerns of households mov-ing in response to state tax policies, Table 16 shows these estimates based on the initialstate in which households resided, ignoring any subsequent cross-state moves.

Table 1 reports results from estimating Equation (2). Column (1) shows that a 65 to 79year olds are 14.8 percentage points less likely to hold a mortgage than those who are 50to 64 years of age. Adding in the relative value of the MID from tax rates in Column (2)shows no overall effect of the value of the MID on mortgage rates. A one percentage pointincrease in the MID increases the rate of mortgage holding by 0.6 percentage points for 65to 79 year olds without changing the mortgage rate for 50 to 64 year olds.17 Given that theaverage combined federal and state mortgage interest marginal subsidy rate has decreasedby 8 percentage points between 1980 and 2015, this implies that changes in the MID mayexplain around 5 points of the relative change in mortgage usage between 50 to 64 year

16The NBER calculates the marginal tax rates for the mortgage interest deduction based on a large, fixed,nationally representative sample of 1995 individual tax returns for each state and year.

17The MID subsidy value ranges from -9.04 to 1.16 with a mean of -2.68, implying a reduction in taxliability when deducting mortgage interest. Lower (more negative) values of the MID variable indicate alarger subsidy. Thus a greater subsidy indicates a greater likelihood of holding a mortgage for 65 to 79 yearolds.

17

olds and 65 to 79 year olds over this time period. There are no statistically significanteffects for the 80 to 100 year old group, suggesting the decrease in the gap between themortgaged rates of those age 80 to 100 and those age 50 to 64 is not explained by these taxincentives. Columns (3)-(4) repeat the prior estimates, but conditional on homeownership.Column (3) reports a larger overall gap in mortgage rates between 65 to 79 year olds and50 to 64 year olds in the subsample of homeowners (0.245). Column (4) suggests an evenlarger effect of the MID on mortgage rates among homeowners than in Column (2) for 65to 79 year olds, about 0.8 percentage points.

Next, we investigate two economic factors that may contribute to older householdsholding onto mortgages longer: local unemployment rates and rent-to-price ratios as es-timated by Campbell et al. (2009). Since Campbell et al. (2009) provides RTP estimatesthrough 2007, we can not include observations beyond 2007 for these specifications. Weassign individuals to RTP data based on their metropolitan area of residence or their cen-sus region if RTP estimates for their metro are are not available. If we instead include onlythe unemployment rate in areas with and without RTP data, our results remain consistent.

Table 2 Column (2) shows that a one percentage point increase in the unemploymentrate decreases the relative likelihood of holding a mortgage by 0.5 percentage points andincreases the likelihood of holding a mortgage for 65 to 79 year olds by 0.5 percentagepoints. Thus, a one unit increase in the unemployment rate decreased the relative mortgagerate gap, by roughly 1 percentage point between those households age 50 to 64 and 65 to79 year olds, and 1.5 percentage points between 50 to 64 and 80 to 100 year olds. This isconsistent with the trend in Figure 4 Panels A and B where the gap in borrowing by agedecreased most during the Great Recession. Given that average unemployment rates in theGreat Recession (2007-2009) were roughly 10 percent (U.S. Bureau of Labor Statistics,2012), or four units higher than in 1999, this increased unemployment decreased the gap by4 points. The estimates conditional on ownership in Column (4) are similar in magnitude.

Table 2 also includes the interaction between age cohorts and the Rent to Price (RTP)ratio, or the ratio of semiannual rents to the purchase price of similar homes (see Campbellet al. (2009) for a comprehensive discussion of this measure). We choose the RTP measureover a traditional home price measure to capture the opportunity cost of owning versusrenting. We use RTP data estimated by Campbell et al. (2009) between 1975 to 2007for 24 metro areas and the four census regions. For interpretation, multiply the RTP ratioestimate by 100. Across the US, between 1994 to 2014 the RTP ratio declines significantlyfrom 5.0 to 3.7. The RTP ratio has little effect on mortgage borrowing, except among the80 to 100 year old group. This suggests that older households in areas with higher rentsrelative to home values are more likely to have a mortgage—perhaps because the transitionto renting would not lower their monthly housing costs by as much as it would in lowerrent areas. The estimates in Column (4) conditional on ownership, are similar.

18

Appendix Tables 13-14 replicate Tables 1-2 but includes household fixed effects inEquation (2). Column (1) shows a decrease in borrowing for those aged 65 to 79 as abaseline with household fixed effects. This means that those that age into a new categorywere 1.7 percentage points less likely to hold a mortgage at age 65 to 79 than age 50 to 64.While the MID effect remains the same in magnitude for 65 to 79 year olds, it is no longerstatistically different from zero in either Columns (2) or (4). We take this as evidence thateven with less power, the signs on the MID effects still remain consistent.

Nearly all coefficients are consistent across Tables 14 and 2, though the overall effectof RTP on mortgage rates is stronger. Since the identification is coming from withinhousehold changes, this could be coming from individuals’ choices to relocate in responseto prices—and differentially by age.

5.4 Are More Recent Elderly Borrowers Different from Prior Co-horts?

Finally, we explore the degree to which mortgage borrowers look different across yearsin the HRS data. We run separate regressions for each age band (age 50 to 64 and 65to 79), including all of the control variables from Equation 1. For simplicity in graph-ing the results, we bundle wealth into two instead of four categories at the median. Thedependent variable in each regression is a mortgage indicator. Figure 9 reports each co-efficient estimate for each age group in each year. If lenders are originating additionalmortgages among older people with certain characteristics, the increase in the mortgagedrate among seniorshttps://www.ssa.gov/policy/docs/rsnotes/rsn2013-01.html compared to50 to 64 year olds may be driven by supply-side factors.

Figure 9 shows that there are no noticeable differences in the influence of income,number of children, education, work status, or marital status across these age groups overtime. The confidence intervals for each of these coefficients overlap, and are very closefor bequest motives.18 There are differences along some dimensions. Those age 65 to 79are more likely to have mortgages with lower levels of wealth in recent years than 50 to 64year olds are. Non-white 65 to 79 year olds are more likely to hold mortgages than whites,but this likelihood has decreased over time to zero. For 50 to 64 year olds, non-whites wereequally likely to hold mortgages as whites in the same age range, and the likelihood hasalso decreased over time. Race explains less of the trend than wealth in the prior figure.

18While bequest motives optically increase the likelihood of having a mortgage for 50 to 64 year olds bymore than they do for 65 to 79 year olds, the magnitude of the association is small–a 0.002 percentage pointincrease in the year with the greatest correlation.

19

5.5 DefaultHigher mortgage debt is a problem if households cannot service the debt payments. Thisis particularly true in retirement, as households have limited ability to counteract financialtroubles with increased labor supply making loan defaults more painful for elderly bor-rowers. The pattern of loan defaults — defined as being 60 days late —relative to otherage cohorts could signal that as older households take on more debt they also are morelikely to default. Figure 10 shows that older households are less likely to be at least 60days late on any account than younger cohorts are in most periods, though there has beena decrease in the gap from 1998 to 2016.

However, overall delinquency rates of the SCF sample are less than 3 percent fornearly every period. Further, the magnitude of the difference across age groups is small:0.045 percentage points in 1998, where the effect size is largest. We interpret this asevidence that older borrowers are unlikely to be risky borrowers—which is consistent withrecent trends in defaults. This is consistent with findings in Brown et al. (2016), showingthat older borrowers (age 50 to 80) have the highest credit scores and lowest delinquencyrates.

Note however, that rates of mortgage borrowing overall have increased considerablyamong older households, and in the aggregate more elderly people will mechanically bein mortgage default. Even though default rates have not substantially increased, and rel-ative to slightly younger age cohorts the rates are not rising, the sheer number of olderhouseholds defaulting may raise some concerns given how severe the consequences ofmortgage foreclosure can be for this group. Elderly cohorts have more exposure to defaultrisk through mortgages, even if relative default rates are not elevated in an economicallymeaningful way in our estimates.

6 DiscussionOlder households are using mortgages at higher rates than prior generations did. This trendis not explained by increasing levels of income or education, or cohort shifts in marriagerates, urban location, or race. Rising wealth does not substantially explain rising borrowingoverall or within age groups over time. Rising rates of homeownership are an importantcomponent of this trend: more homeowners mechanically means more households whocan borrow against home equity. Indeed, mortgage borrowing among older householdsaccelerated with the housing boom in the mid-2000s. Changes to tax laws also appear tocontribute to differential changes in mortgage usage by age.

We find each percentage point increase in the subsidy for mortgage interest increasesthe likelihood of a 65 to 79 year old household having a mortgage by roughly 0.6 per-

20

centage points. According to NBER estimates, the average combined state and federalmarginal subsidy rate for mortgage interest increased by 8 percentage points between 1980and 2014. We estimate approximately 5 percentage points of the growth in mortgage usageby this people age 65 to 79 compared to 50 to 64 year olds can be attributed to changingtax policy.

Relatively older homeowners without other assets, especially non-retirement assets,may simply be borrowing to fund consumption in the present—there are some patterns ofborrowing in response to local unemployment rates that are consistent with this concept.This could be direct consumption, or to help family members.

Many of these households are actively originating new mortgages or increasing mort-gage balances through cash-out refinances. In the HRS data, we find that nine percent ofhomeowners over age 65 originate new mortgages or increase their mortgage balance by atleast $5,000. Among older renters who transition to homeownership, half use a mortgageto finance the purchase.

Low interest rates and rising home values may have enticed many homeowners intheir 50s and 60s into refinancing in the 2000s. Given the decline in home equity in thelate 2000s, paying off these loans may not have been feasible. Whether these patterns aresustained as housing prices appreciate remains to be seen.

Adults age 62 and older have access to the federally insured Home Equity ConversionMortgage (HECM) reverse mortgage. These loans allow older adults to consume fromthe equity in their homes without any monthly repayment. Just over one million HECMloans have been originated in the last two decades, representing only about two percentof the eligible population with a mortgage (Moulton et al., 2017). The lack of use of thisprogram may deserve more attention, given the need for liquidity that households haveand the potential to tap equity without a monthly payment. For example, studies like thoseby Davidoff and colleagues (Davidoff et al., 2017), may point to a lack of understandingof HECMs and how they could be useful.

Policymakers managing social insurance obligations have an interest in facilitatinghouseholds to unlock home equity to fund their consumption and reduce reliance in publicinsurance programs (French et al., 2018). The rise of mortgage use by older householdsis one way that people can tap their home equity, borrowing against their real propertyto pay for living expenses instead of drawing benefits. The standard forward mortgage isubiquitous and efficient, serving as an alternative to a more complicated reverse mortgage.Of course, households have to be able to keep up with monthly payments for this strategyto be useful.

The increase in the use of mortgages by older households is an important issue forlenders and financial institutions to monitor going forward as more cohorts of older house-holds retire, and existing retirees either take on more debt or pay off their loans. Likewise,

21

estate sales of property and probate courts may find more homes encumbered with a mort-gage. Surviving widows and widowers may struggle to pay mortgage payments after thedeath of a spouse and may face a reduction of pension or Social Security payments. Thismay be a form of default risk not currently priced into mortgage underwriting for olderloan applicants. If more mortgage borrowing among the elderly results in more foreclo-sures, smaller inheritances, or even estates with negative values, this could have negativeeffects on extended families and communities.

While researchers have mixed views of debt overall, there may be insights to drawfrom ongoing work in fields such as sociology. For example, work by scholars like Dwyer(2018) focus on use of debt by race and class, and how the use of different kinds of debtin differing contexts may lead to growing economic inequality. Another potential areato study is how people have developed preferences for debt over time. The Baby Boomgeneration, who will soon make up the majority of the elderly population, are the first gen-eration to have grown up with easy access to credit, and a 30-year mortgage market. Theyalso experienced high interest rates in the 80s, and may view debt as relatively cheap today.This is a generation with high levels of wealth, education and income, as well as healthand longevity (Guido et al., 2018). Understanding the influences and preference formationrelated to credit use for this unique age cohort may help further our understanding of thisgeneration’s housing and mortgage choices.

ReferencesAmromin, Gene, Jennifer Huang, and Clemens Sialm, “The tradeoff between mortgage

prepayments and tax-deferred retirement savings,” Journal of Public Economics, 2007,91 (10), 2014–2040.

Argys, Laura M, Andrew Friedson, and M Melinda Pitts, “Killer Debt: The Impact ofDebt on Mortality,” FRB Atlanta Working Paper, 2016, (No. 2016-14).

Bedard, Nicolas, Pierre-Carl Michaud et al., Playing with Fire?: Debt Near Retirementin Canada, Center for Interuniversity Research and Analysis on Organizations, 2018.

Bhutta, Neil and Benjamin J Keys, “Interest rates and equity extraction during the hous-ing boom,” The American Economic Review, 2016, 106 (7), 1742–1774.

Bian, Xun, “Leverage and Elderly Homeowners’ Decisions to Downsize,” Housing Stud-ies, Apr 2015, 31 (1), 20–41.

22

Bricker, Jesse, Brian Bucks, Arthur Kennickell, Traci Mach, and Kevin Moore,“The financial crisis from the family’s perspective: Evidence from the 2007–2009 SCFpanel,” Journal of Consumer Affairs, 2012, 46 (3), 537–555.

, Lisa J Dettling, Alice Henriques, Joanne W Hsu, Lindsay Jacobs, Kevin B Moore,Sarah Pack, John Sabelhaus, Jeffrey Thompson, and Richard A Windle, “Changesin US family finances from 2013 to 2016: evidence from the survey of consumer fi-nances,” Fed. Res. Bull., 2017, 103, 1.

Brown, Jason, Karen Dynan, and Theodore Figinski, “The Risk of Financial Hard-ship in Retirement: A Cohort Analysis,” in “in” Wharton Pension Research CouncilSymposium 2019.

Brown, Meta, Donghoon Lee, Joelle Scally, Katherine Strair, and H. Wilbert Vander Klaauw, “The Graying of American Debt,” Federal Reserve Bank of New York,Research and Statistics Group, Microeconomic Studies, February, 2016, February 24,2016–9.

, Sarah Stein, and Basit Zafar, “The impact of housing markets on consumer debt:credit report evidence from 1999 to 2012,” Journal of Money, Credit, and Banking,2015, 47 (S1), 175–213.

Brule, Gael and Laura Ravazzini, “The Four Forms of Wealth and Happiness: HowDifferent Forms of Wealth Affect the Subjective Well-being of the Elderly in Europe,”Wealth(s) and Subjective Well-Being, 2019, pp. 199–219.

Burman, Leonard E, Christopher Geissler, and Eric J Toder, “How big are total indi-vidual income tax expenditures, and who benefits from them?,” The American EconomicReview, 2008, 98 (2), 79–83.

Butrica, Barbara A and Nadia S Karamcheva, “In Debt and Approaching Retirement:Claim Social Security or Work Longer?,” AEA Papers and Proceedings, 2018, 108,401–06.

Campbell, Sean D, Morris A Davis, Joshua Gallin, and Robert F Martin, “Whatmoves housing markets: A variance decomposition of the rent–price ratio,” Journalof Urban Economics, 2009, 66 (2), 90–102.

Consumer Financial Protection Bureau, “Snapshot of older consumers and mortgagedebt,” CFPB Office for Older Americans, 2014, May, 1–18.

23

Copeland, Craig, “Debt of the Elderly and Near Elderly, 1992-2013,” EBRI Notes, 2015,36 (1).

Davidoff, Thomas, “Home equity commitment and long-term care insurance demand,”Journal of Public Economics, 2010, 94, 44–49.

, Patrick Gerhard, and Thomas Post, “Reverse mortgages: What homeowners (don’t)know and how it matters,” Journal of Economic Behavior & Organization, 2017, 133,151–171.

DeFusco, Anthony A and Andrew Paciorek, “The interest rate elasticity of mortgagedemand: Evidence from bunching at the conforming loan limit,” American EconomicJournal: Economic Policy, 2017, 9 (1), 210–40.

Dietz, Robert D and Donald R Haurin, “The social and private micro-level conse-quences of homeownership,” Journal of Urban Economics, 2003, 54 (3), 401–450.

Drentea, Patricia and John R Reynolds, “Where does debt fit in the stress processmodel?,” Society and Mental Health, 2015, 5 (1), 16–32.

Dunn, Lucia F and Ida A Mirzaie, “Consumer debt stress, changes in household debt,and the Great Recession,” Economic Inquiry, 2016, 54 (1), 201–214.

Dunsky, Robert and James Follain, “Tax-Induced Portfolio Reshuffling: The Case ofthe Mortgage Interest Deduction,” Real Estate Economics, 2000, 28, 638–718.

Dwyer, Rachel E, “Credit, debt, and inequality,” Annual Review of Sociology, 2018, 44,237–261.

Engelhardt, Gary, Michael Eriksen, William Gale, and Gregory Mills, “What Are theSocial Benefits of Homeownership? Experimental Evidence for Low-Income House-holds,” Journal of Urban Economics, 2010, 67, 249–258.

Engelhardt, Gary V., “Social security and elderly homeownership,” Journal of UrbanEconomics, 2008, 63 (1), 280 – 305.

and Nadia Greenhalgh-Stanley, “Home health care and the housing and living ar-rangements of the elderly,” Journal of Urban Economics, 2010, 67 (2), 226 – 238.

French, Declan, Donal McKillop, and Tripti Sharma, “What determines UK housingequity withdrawal in later life?,” Regional Science and Urban Economics, 2018, 73,143–154.

24

Gale, William, Jonathan Gruber, and Seth Stepehns-Davidowtiz, “EncouragingHomeownership Through the Tax Code,” Tax Notes, 2007, 115, 1171–1189.

Golant, Stephen M, “Commentary: Irrational exuberance for the aging in place of vul-nerable low-income older homeowners,” Journal of Aging & Social Policy, 2008, 20(4), 379–397.

, “Low-income Elderly Homeowners in Very Old Dwellings: the Need for Public PolicyDebate,” Journal of Aging & Social Policy, 2008, 20 (1), 1–28.

Goodman, Laurie S and Christopher Mayer, “Homeownership and the AmericanDream,” Journal of Economic Perspectives, 2018, 32 (1), 31–58.

Gruber, Jonathan, Amalie Jensen, and Henrik Kleven, “Do People Respond to theMortage Interest Deduction? Quasi-Experimental Evidence from Denmark,” NBERWorking Paper, 2017, 23600.

Guido, Gianluigi, Marco Pichierri, Giovanni Pino, and Roberta Conoci, “The seg-mentation of elderly consumers: A literature review,” Journal of Customer Behaviour,2018, 17 (4), 257–278.

Gyourko, Joseph and Joseph Tracy, “Using Home Maintenance and Repairs to SmoothVariable Earnings,” The Review of Economics and Statistics, 2006, 88 (4), 736–747.

Hanson, Andrew, “Size of home, homeownership, and the mortgage interest deduction,”Journal of Housing Economics, 2012, 21 (3), 195–210.

Hilber, Christian and Tracy Turner, “The Mortgage Interest Deduction and Its Impacton Homeownership Decisions,” Review of Economics and Statistics, 2014, 96 (4), 618–637.

Jappelli, T., M. Pagano, and M. Di Maggio, “Households’ indebtedness and financialfragility,” Journal of Financial Management, Markets and Institutions, 2013, 1 (1), 26–35.

Jappelli, Tullio and Luigi Pistaferri, “Do people respond to tax incentives? An anal-ysis of the Italian reform of the deductibility of home mortgage interests,” EuropeanEconomic Review, 2007, 51 (2), 26–35.

Ling, David and Gary McGill, “Evidence on the Demand for Mortgage Debt by Owner-Occupants,” Journal of Urban Economics, 1998, 44, 391–414.

25

Lusardi, Annamaria and Olivia Mitchell, “Baby boomer retirement security: The rolesof planning, financial literacy, and housing wealth.,” Journal of Monetary Economics,2007, 54 (1), 205–224.

, , and Noemi Oggero, “Debt and Financial Vulnerability on the Verge of Retire-ment,” NBER Working Paper, 2017, August (23664).

, Olivia S Mitchell, and Noemi Oggero, “The changing face of debt and financialfragility at older ages,” AEA Papers and Proceedings, 2018, 108, 407–11.

Mayer, Christopher J, “Housing, Mortgages, and Retirement,” Evidence and Innovationin Housing Law and Policy, 2017, p. 203.

Moulton, Stephanie, Cazilia Loibl, and Donald Haurin, “Reverse mortgage motivationsand outcomes: Insights from survey data,” Cityscape, 2017, 19 (1), 73.

, Donald Haurin, and Caezilia Loibl, “Debt Stress and Mortgage Borrowing in OlderAge: Implications for Economic Security in Retirement,” CFS RDRC Working Paper,University of Wisconsin, 2019, 2019-06.

Nason, Deborah, “Pros and cons of paying off mortgage before retirement,” CNBC, 2017,22 June 2017.

Painter, Gary and KwanOk Lee, “Housing tenure transitions of older households: Lifecycle, demographic, and familial factors,” Regional Science and Urban Economics,2009, 39 (6), 749–760.

Pham, Sherisse, “Retirements swallowed by debt,” New York Times, 2011.

Poterba, James and Todd Sinai, “Tax expenditures for owner-occupied Housing: De-ductions for Property Taxes and mortgage interest and the exclusion of imputed rentalincome,” American Economic Review, 2008, 98 (2), 84–89.

Poterba, James M, “Retirement security in an aging population,” The American Eco-nomic Review, 2014, 104 (5), 1–30.

Poterba, James, Steven Venti, and David Wise, “The composition and drawdown ofwealth in retirement,” The Journal of Economic Perspectives, 2011, 25 (4), 95–117.

Sabia, Joseph J, “There’s no place like home: a hazard model analysis of aging in placeamong older homeowners in the PSID,” Research on Aging, 2008, 30 (1), 3–35.

26

Shan, Hui, “Property taxes and elderly mobility,” Journal of Urban Economics, 2010, 67(2), 194 – 205.

, “Reversing the trend The recent expansion of the reverse mortgage market,” Real Es-tate Economics, 2011, 39 (4), 743–768.

Sinai, Todd and Nicholas S Souleles, “Owner-occupied housing as a hedge against rentrisk,” The Quarterly Journal of Economics, 2005, 120 (2), 763–789.

Solon, Gary, Steven J Haider, and Jeffrey M Wooldridge, “What are we weightingfor?,” Journal of Human Resources, 2015, 50 (2), 301–316.

Sommer, Kamila and Paul Sullivan, “Implications of US tax policy for house prices,rents, and homeownership,” American Economic Review, 2018, 108 (2), 241–74.

U.S. Bureau of Labor Statistics, “The Recession of 2007-2009,” BLS Spotlight on Statis-tics, 2012.

Venti, Steven F. and David A. Wise, “Aging and the income value of housing wealth,”Journal of Public Economics, 1991, 44 (3), 371 – 397. Special Issue on InternationalStudies of Housing Demand: The Effects of Government Policies.

Vornovytskyy, Marina, Alfred Gottschalck, and Adam Smith, “Household Debt in theUS: 2000 to 2011,” Technical Report, US Census Bureau, Washington, DC 2011.

Zhao, Lingxiao and Gregory Burge, “Housing Wealth, Property Taxes, and Labor Sup-ply among the Elderly,” Journal of Labor Economics, Jan 2017, 35 (1), 227–263.

Zinman, Jonathan, “Household Debt: Facts, Puzzles, Theories, and Policies,” AnnualReview of Economics, Aug 2015, 7 (1), 251–276.

27

Figure 1: Mortgage Holding Rates and Homeownership by Age Profiles over Time

Panel A: Mortgaged Rate Panel B: Homeownership0

.2.4

.6M

ortg

aged

Rat

e

20 40 60 80 100Age

1980 1990 2000 2015

.2.4

.6.8

Hom

eow

ners

hip

Rat

e

20 40 60 80 100Age

1980 1990 2000 2015

Source: Census PUMS Data 1980-2000, 2015 American Community Survey Data.

Figure 2: Wealth & Retirement Age Profiles

Panel A: Wealth Panel B: Retirement

100

200

300

400

500

Wea

lth (0

00s,

$20

14)

50 60 70 80 90Age

1998-2002 2004-2008 2010-2014

0.2

.4.6

.81

Frac

tion

Ret

ired

50 60 70 80 90Age

1998-2002 2004-2008 2010-2014

Source: Health and Retirement Study.

28

Figure 3: Wealth by Age by Time Period within Quintile

20th Wealth Percentile

010

2030

4050

Wea

lth (0

00s,

$20

14)

50 60 70 80 90 100Age

1998-2002 2004-2008 2010-2014

50th Wealth Percentile

010

020

030

0W

ealth

(000

s, $

2014

)

50 60 70 80 90 100Age

1998-2002 2004-2008 2010-2014

80th Wealth Percentile

020

040

060

080

0W

ealth

(000

s, $

2014

)

50 60 70 80 90 100Age

1998-2002 2004-2008 2010-2014

Source: Health and Retirement Study.

29

Figure 4: The Prevalence of Mortgage by Age Cohort over Time

A. Survey of Consumer FinancesDifference: 65-79 and 50-64 year olds Difference: 80-100 and 50-64 year olds

B. Health and Retirement StudyDifference: 65-79 and 50-64 year olds Difference: 80-100 and 50-64 year olds

Estimates α1 and α2 coefficients from Equation 1 in each dataset with 95% confidence intervals reported foreach coefficient. SCF estimates are in Appendix Table 3. HRS estimates are in Appendix Table 4. Controlsinclude number of children, logged income, greater than high school education, employment status, networth quartiles, race, and marital status. HRS controls additionally include self-reported health status andbequest motive variables.

30

Figure 5: Homeownership Rates Increasing for 65-79 Relative to 50-64 Year Olds

A. Homeownership B. Mortgages Among Homeowners

Source: Survey of Consumer Finances. We report estimates of α1 coefficients from Equation 1 with 95%confidence intervals. All estimates are in Appendix Tables 5 and 6. Controls include number of children,logged income, greater than high school education, employment status, net worth quartiles, race, and mari-tal status.

31

Figure 6: Home Values and Equity Extraction for 65-79 Relative to 50-64 Year Olds

A. Home Value B. Home Equity

C. Extract Equity D. Refinance

Source: Survey of Consumer Finances. We report estimates of α1 coefficients from Equation 1 with 95%confidence intervals. Coefficient estimates are in Appendix Tables 7 and 8. Controls include number ofchildren, logged income, greater than high school education, employment status, net worth quartiles, race,and marital status.

32

Figure 7: Heterogeneity in Age Profiles

A. Bequest Motive B. Pension

C. Education Level D. Wealth

Source: Health and Retirement Study data in panels A and D; 1989-2016 Survey of Consumer Finances datain panels B and C. We report estimates of α1 coefficients from Equation 1 with 95% confidence intervals.Coefficient estimates are in Appendix Tables 9-12. Controls include number of children, logged income,greater than high school education, race, employment status, net worth quartiles, and marital status.

33

Figure 8: Heterogeneity in Age Profiles, homeowners only

A. Bequest Motive B. Pension

C. Education Level D. Wealth

Source: Health and Retirement Study data in panels A and D; 1989-2016 Survey of Consumer Finances datain panels B and C. We report estimates of α1 coefficients from Equation 1 with 95% confidence intervals.Coefficient estimates are in Appendix Tables 9-12. Controls include number of children, logged income,greater than high school education, race, employment status, net worth quartiles, and marital status.

34

Table 1: Mortgage Interest Tax Treatment and Mortgage Rate

(1) (2) (3) (4)Base + MID Base + MID

Age 65-79 -0.148*** -0.146*** -0.254*** -0.243***(0.004) (0.007) (0.005) (0.009)

Age 80-100 -0.250*** -0.241*** -0.340*** -0.307***(0.005) (0.009) (0.007) (0.012)

65-79 X MID -0.006*** -0.008***(0.002) (0.003)

80-100 X MID 0.003 -0.000(0.003) (0.004)

MID 0.000 -0.005(0.003) (0.004)

Always Own X XMSA FE X X X XYear FE X X X XControls X X X XObservations 71,327 71,327 41,063 41,063* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Each observation is a respondent-year. All models include MSA andyear fixed effects and control for number of children, logged income, top state marginal income tax rate,urban area, greater than high school education, race, marital status, bequest motive, net worth quartiles,and subjective health measure. MID is the average marginal mortgage interest tax rate by state and yearfrom NBER.

35

Table 2: Mortgage Rate by State Unemployment and MSA Rent to Price Ratio (RTP)

(1) (2) (3) (4)Base +UE,RTP Base +UE, RTP

Age 65-79 -0.166*** -0.200*** -0.230*** -0.311***(0.005) (0.028) (0.007) (0.040)

Age 80-100 -0.261*** -0.373*** -0.297*** -0.580***(0.007) (0.037) (0.010) (0.056)

65-79 X Unemployment 0.010*** 0.008**(0.002) (0.003)

80-100 X Unemployment 0.015*** 0.020***(0.003) (0.005)

65-79 X RTP -0.004 0.009(0.005) (0.007)

80-100 X RTP 0.008 0.040***(0.007) (0.010)

Unemployment -0.005** -0.008***(0.002) (0.003)

RTP 0.005 0.001(0.012) (0.017)

Always Own X XMSA FE X X X XYear FE X X X XControls X X X XObservations 42,927 42,927 25,009 25,009* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Each observation is a respondent-year. All models include MSA andyear fixed effects and control for number of children, logged income, urban area, greater than high schooleducation, race, marital status, bequest motive, net worth quartiles, and subjective health measure.Unemployment rate is the state unemployment rate that year. RTP is rent to price ratio as estimated byCampbell et al. (2009).

36

Figure 9: Mortgage Rates by Age Cohort by Demographic Factor

ln(Income) # Children

.02

.04

.06

.08

.1.1

2C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

-.01

0.0

1.0

2C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

Bequest Education

0.0

01.0

02.0

03C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

-.15

-.1-.0

50

Coe

ffici

ent E

stim

ate

1995 2000 2005 2010 2015

50-64 65-79

Work Married

-.05

0.0

5.1

.15

.2C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

-.50

.5C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

Wealth Below Median Non-White

-.2-.1

0.1

.2.3

Coe

ffici

ent E

stim

ate

1995 2000 2005 2010 2015

50-64 65-79

-.1-.0

50

.05

.1.1

5C

oeffi

cien

t Est

imat

e

1995 2000 2005 2010 2015

50-64 65-79

Source: Health and Retirement Study. This model’s dependent variable is a mortgage indicator and theindependent variable is each control variable. We run a separate regression for each age category 50-64and 65-79 for each year. We report the coefficient estimates for each age group and year, as well as the 95%confidence interval.

37

Figure 10: Late Payments for 65-79 Relative to 50-64 Year Olds

Source: Survey of Consumer Finances. We report estimates of α1 coefficients from Equation 1 with 95%confidence intervals. Coefficient estimates are in Appendix Table 15.

38

7 Appendix

Figure 11: Homeownership Rates Increasing for Older Households Relative to 50-64 YearOlds

CensusHomeownership Mortgages Among Homeowners

Source: Census Data. We report estimates of α1 coefficients from Equation 1 with 95% confidence intervals.

39

Table 3: Mortgage Rates by Age Cohort over Time, SCF

1989 1992 1995 1998 2001 2004 2007 2010 2013 2016Panel A: SCF, No Controls65-79 -0.265*** -0.307*** -0.315*** -0.307*** -0.269*** -0.246*** -0.219*** -0.186*** -0.136*** -0.129***

(0.0139) (0.0125) (0.0127) (0.0128) (0.0124) (0.0126) (0.0130) (0.00987) (0.00996) (0.00951)80-100 -0.426*** -0.425*** -0.466*** -0.450*** -0.477*** -0.472*** -0.521*** -0.371*** -0.350*** -0.290***

(0.0124) (0.0138) (0.0134) (0.0137) (0.0119) (0.0131) (0.0119) (0.0132) (0.0127) (0.0132)N 8030 9270 10140 10329 10661 11736 12040 17050 16745 18210

Panel B: SCF, Controls65-79 -0.112*** -0.162*** -0.206*** -0.165*** -0.136*** -0.120*** -0.0983*** -0.0842*** -0.0474*** -0.0241**

(0.0153) (0.0153) (0.0150) (0.0150) (0.0153) (0.0141) (0.0147) (0.0108) (0.0108) (0.0105)80-100 -0.230*** -0.242*** -0.305*** -0.239*** -0.263*** -0.284*** -0.316*** -0.199*** -0.170*** -0.144***

(0.0162) (0.0171) (0.0165) (0.0178) (0.0171) (0.0165) (0.0166) (0.0156) (0.0148) (0.0152)N 7792 8881 9700 10058 10407 11320 11658 16389 16171 17569

Notes: Survey of Consumer Finances. Controls include race, marital status, and high school education indicators, as well as logged

income, and number of children. Household weights are used in all specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05,

*** p<0.01

40

Table 4: Mortgage Rates by Age Cohort over Time, HRS

1998 2000 2002 2004 2006 2008 2010 2012 2014Panel A: HRS, No Controls65-79 -0.297*** -0.259*** -0.230*** -0.246*** -0.215*** -0.211*** -0.199*** -0.175*** -0.143***

(0.012) (0.014) (0.015) (0.013) (0.015) (0.016) (0.014) (0.015) (0.017)80-100 -0.425*** -0.385*** -0.364*** -0.401*** -0.401*** -0.399*** -0.389*** -0.362*** -0.331***

(0.011) (0.013) (0.014) (0.013) (0.014) (0.015) (0.013) (0.014) (0.015)N 9077 7409 6691 7267 6558 6013 8357 7593 6974Panel B: HRS, Controls65-79 -0.166*** -0.145*** -0.149*** -0.169*** -0.125*** -0.147*** -0.162*** -0.134*** -0.114***

(0.013) (0.015) (0.016) (0.015) (0.016) (0.018) (0.017) (0.017) (0.018)80-100 -0.231*** -0.213*** -0.227*** -0.277*** -0.253*** -0.267*** -0.283*** -0.252*** -0.239***

(0.015) (0.016) (0.017) (0.017) (0.018) (0.019) (0.017) (0.018) (0.019)N 8186 7053 6443 7007 6375 5853 8026 7448 6847

Notes: Source: Health and Retirement Study. Controls include race, marital status, and high school education indicators, as well as

logged income, and number of children. Household weights are used in all specifications. Excluded age group is 50-64. * p<0.10, **

p<0.05, *** p<0.01

Table 5: Homeownership Rates by Age Cohort over Time, SCF

1989 1992 1995 1998 2001 2004 2007 2010 2013 2016Panel A: Homeownership65-79 -0.0391*** 0.0253** -0.0170 0.0208** 0.0154 0.0466*** 0.0433*** 0.0553*** 0.114*** 0.0804***

(0.0139) (0.0115) (0.0118) (0.0105) (0.00973) (0.00981) (0.00966) (0.00758) (0.00776) (0.00779)80-100 -0.143*** -0.0851*** -0.156*** -0.0223 -0.104*** 0.0105 -0.0733*** 0.0276** 0.0633*** 0.0786***

(0.0240) (0.0209) (0.0224) (0.0164) (0.0175) (0.0160) (0.0161) (0.0124) (0.0129) (0.0121)N 8030 9270 10140 10329 10661 11736 12040 17050 16745 18210

Panel B: Homeownership, Controls65-79 -0.0263** 0.0204* 0.0144 0.0394*** 0.0112 0.0635*** 0.0691*** 0.0586*** 0.142*** 0.0948***

(0.0124) (0.0122) (0.0110) (0.0104) (0.0109) (0.00952) (0.00917) (0.00718) (0.00707) (0.00738)80-100 -0.125*** -0.0849*** -0.0969*** 0.00930 -0.113*** 0.0442*** -0.0324** 0.0416*** 0.116*** 0.0980***

(0.0201) (0.0195) (0.0207) (0.0161) (0.0165) (0.0152) (0.0136) (0.0111) (0.0113) (0.0114)N 7,792 8,881 9,700 10,058 10,407 11,320 11,658 16,389 16,171 17,569

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 5. All models include controls for race, marital status,

and high school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights are

used in all specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

41

Table 6: Mortgage Rates Conditional on Homeownership by Age Cohort over Time

1989 1992 1995 1998 2001 2004 2007 2010 2013 2016

Panel A: Mortgage Conditional on Homeownership65-79 -0.319*** -0.399*** -0.383*** -0.399*** -0.339*** -0.331*** -0.295*** -0.273*** -0.255*** -0.228***

(0.0161) (0.0144) (0.0144) (0.0147) (0.0141) (0.0138) (0.0140) (0.0109) (0.0110) (0.0107)80-100 -0.524*** -0.535*** -0.562*** -0.571*** -0.581*** -0.594*** -0.635*** -0.487*** -0.496*** -0.427***

(0.0152) (0.0176) (0.0176) (0.0167) (0.0150) (0.0153) (0.0145) (0.0156) (0.0152) (0.0156)N 6795 7865 8855 8591 8917 10084 10350 13835 13540 14685

Panel B: Mortgage Conditional on Homeownership, Controls65-79 -0.118*** -0.218*** -0.253*** -0.226*** -0.165*** -0.176*** -0.167*** -0.152*** -0.165*** -0.102***

(0.0195) (0.0185) (0.0176) (0.0180) (0.0177) (0.0165) (0.0164) (0.0121) (0.0124) (0.0126)80-100 -0.273*** -0.299*** -0.366*** -0.319*** -0.290*** -0.378*** -0.395*** -0.282*** -0.297*** -0.255***

(0.0206) (0.0224) (0.0211) (0.0225) (0.0216) (0.0202) (0.0206) (0.0184) (0.0183) (0.0177)N 6587 7521 8475 8405 8718 9780 10043 13339 13111 14157

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 5. All models include controls for race, marital status,

and high school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights are

used in all specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

42

Table 7: Home Values and Equity Conditional on Homeownership by Age Cohort overTime

1989 1992 1995 1998 2001 2004 2007 2010 2013 2016

Panel A1: Home Value Conditional on Homeownership65-79 -70588*** -58623*** -47039*** -51696*** -33013*** -114067*** -28961*** -45136*** -15139*** -42669***

(10076) (7649) (5853) (7265) (7559) (8115) (8816) (6575) (5587) (5358)80-100 -82258*** -91922*** -68106*** -105492*** -78866*** -86050*** -108383*** -93640*** -84474*** -38803***

(17670) (12379) (9509) (8658) (10898) (16854) (10461) (8805) (6762) (9042)N 6795 7865 8855 8591 8917 10084 10350 13820 13530 14680

Panel A2: Home Value Conditional on Homeownership, Controls65-79 -2825 -4011 -15544*** 1160 27774*** -32018*** 37553*** 16420*** 36561*** 14013***

(9206) (7727) (5710) (7138) (7473) (8715) (7790) (5194) (5571.4) (5069)80-100 -20441 -28083** -34716*** -34951*** -9343 12269 10117 -121.9 17968** 40648***

(16776) (13295) (8060) (9254) (11395) (15898) (10875) (7817) (7565) (8283)N 6587 7521 8475 8405 8718 9780 10043 13324 13101 14152

Panel B1: Home Equity Conditional on Homeownership65-79 18027*** 27804*** 32954*** -28887*** 37987*** 20084*** 39957*** 10402**

(5196) (6661) (6849) (7288) (7830) (5848) (4819) (4698)80-100 15285* -6359 14888 26149 9561 3927 8286 42319***

(9179) (8162) (10639) (16529) (9977) (8088) (6399) (8641)N 8855 8591 8917 10084 10350 13835 13540 14685

Panel B2: Home Equity Conditional on Homeownership, Controls65-79 25304*** 45210*** 50141*** 5159 60026*** 43383*** 63047*** 34452***

(5029) (6952) (6691) (7651) (6915) (4557) (4882) (4799)80-100 15499** 15466* 17386* 58807*** 45401*** 38772*** 52847*** 79443***

(7273) (8909) (10504) (15249) (10041) (7173) (6959) (8191)N 8475 8405 8718 9780 10043 13339 13111 14157

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 6. Controls include race, marital status, and high

school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights are used in all

specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

43

Table 8: Home Equity Extraction and Refinance Conditional on Homeownership by AgeCohort over Time

1995 1998 2001 2004 2007 2010 2013 2016

Panel A1: Equity Extraction Conditional on Homeownership65-79 -0.0437*** -0.0630*** -0.0616*** -0.103*** -0.0801*** -0.0554*** -0.0286*** -0.0340***

(0.00646) (0.00989) (0.00879) (0.00919) (0.00964) (0.00723) (0.00673) (0.00682)80-100 -0.0728*** -0.134*** -0.119*** -0.156*** -0.159*** -0.0867*** -0.0649*** -0.0745***

(0.00525) (0.00731) (0.00822) (0.00876) (0.00825) (0.00971) (0.00873) (0.00845)N 8855 8591 8917 10084 10350 13835 13540 14685

Panel A2: Equity Extraction Conditional on Homeownership, Controls65-79 -0.0188** -0.0396*** -0.0190* -0.0381*** -0.00701 -0.0283*** -0.0180** -0.0200**

(0.00775) (0.0123) (0.0106) (0.0112) (0.0112) (0.00919) (0.00815) (0.00834)80-100 -0.0315*** -0.0917*** -0.0550*** -0.0691*** -0.0280** -0.0536*** -0.0487*** -0.0627***

(0.00670) (0.0118) (0.0117) (0.0117) (0.0117) (0.0120) (0.0108) (0.0105)N 8475 8405 8718 9780 10043 13339 13111 14157

Panel B1: Refinance Conditional on Homeownership65-79 -0.168*** -0.218*** -0.185*** -0.256*** -0.218*** -0.186*** -0.166*** -0.131***

(0.0105) (0.0127) (0.0121) (0.0125) (0.0129) (0.0103) (0.0108) (0.0105)80-100 -0.232*** -0.320*** -0.311*** -0.375*** -0.382*** -0.306*** -0.326*** -0.332***

(0.0106) (0.0110) (0.0108) (0.0132) (0.0125) (0.0136) (0.0140) (0.0120)N 8855 8591 8917 10084 10350 13835 13540 14685

Panel B2: Refinance Conditional on Homeownership, Controls65-79 -0.114*** -0.130*** -0.0628*** -0.102*** -0.0829*** -0.0969*** -0.0980*** -0.0201

(0.0124) (0.0156) (0.0153) (0.0144) (0.0156) (0.0120) (0.0126) (0.0126)80-100 -0.148*** -0.186*** -0.132*** -0.155*** -0.153*** -0.166*** -0.189*** -0.193***

(0.0132) (0.0158) (0.0165) (0.0169) (0.0179) (0.0163) (0.0171) (0.0149)N 8475 8405 8718 9780 10043 13339 13111 14157

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 6. Controls include race, marital status, and high

school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights are used in all

specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

44

Table 9: HRS: Bequest Heterogeneity in Mortgage Rate by Age Cohort over Time

(1) (2) (3) (4) (5) (6) (7) (8) (9)1998 2000 2002 2004 2006 2008 2010 2012 2014

Panel A1: Mortgage Conditional on Bequest Motive65-79 -0.192*** -0.169*** -0.171*** -0.195*** -0.159*** -0.166*** -0.182*** -0.137*** -0.132***

(0.016) (0.018) (0.018) (0.018) (0.020) (0.021) (0.019) (0.020) (0.021)80-100 -0.257*** -0.246*** -0.253*** -0.320*** -0.307*** -0.311*** -0.320*** -0.277*** -0.282***

(0.017) (0.019) (0.020) (0.020) (0.022) (0.023) (0.020) (0.021) (0.021)N 7090 5726 5207 5592 5027 4650 6382 5883 5408

Panel B1: Mortgage Conditional on No Bequest Motive65-79 -0.042** -0.055** -0.031 -0.050** -0.007 -0.045 -0.030 -0.054* 0.002

(0.020) (0.022) (0.026) (0.022) (0.022) (0.028) (0.026) (0.027) (0.030)80-100 -0.113*** -0.118*** -0.113*** -0.116*** -0.086*** -0.113*** -0.119*** -0.109*** -0.044

(0.021) (0.023) (0.026) (0.023) (0.022) (0.028) (0.025) (0.026) (0.030)N 1910 1619 1411 1552 1415 1247 1786 1608 1466

Panel A2: Mortgage Conditional on Has Bequest Motive, Homeowners65-79 -0.227*** -0.191*** -0.197*** -0.216*** -0.174*** -0.177*** -0.255*** -0.227*** -0.226***

(0.018) (0.020) (0.020) (0.020) (0.022) (0.023) (0.021) (0.022) (0.023)80-100 -0.283*** -0.243*** -0.255*** -0.329*** -0.318*** -0.318*** -0.401*** -0.379*** -0.373***

(0.021) (0.022) (0.024) (0.023) (0.026) (0.027) (0.024) (0.024) (0.025)N 5804 4676 4260 4558 4086 3819 4780 4399 4031

Panel B2: Mortgage Conditional on No Bequest Motive, Homeowners65-79 -0.148*** -0.130** -0.089 -0.179*** -0.044 -0.163** -0.284*** -0.375*** -0.204***

(0.053) (0.059) (0.062) (0.065) (0.068) (0.078) (0.058) (0.067) (0.068)80-100 -0.221*** -0.180*** -0.167** -0.242*** -0.226*** -0.251*** -0.520*** -0.467*** -0.233***

(0.061) (0.064) (0.067) (0.070) (0.074) (0.089) (0.066) (0.069) (0.080)N 658 549 476 477 411 367 474 415 407

Notes: Health and Retirement Study. This table includes the estimates for Figure 7. All models include state fixed effects, as well as

controls include number of children, logged income, greater than high school education, race, and marital status. Excluded age group

is 50-64. Household weights are used in all specifications. * p<0.10, ** p<0.05, *** p<0.01

45

Table 10: SCF: Pension Heterogeneity in Mortgage Rate by Age Cohort and PensionHolding over Time

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)1989 1992 1995 1998 2001 2004 2007 2010 2013 2016

Panel A1: Mortgage Conditional on Pension65-79 -0.103*** -0.218*** -0.245*** -0.140*** -0.156*** -0.144*** -0.130*** -0.0714*** -0.0718*** -0.0646***

(0.0210) (0.0215) (0.0204) (0.0214) (0.0210) (0.0189) (0.0195) (0.0145) (0.0145) (0.0144)80-100 -0.252*** -0.276*** -0.336*** -0.235*** -0.246*** -0.371*** -0.338*** -0.242*** -0.224*** -0.169***

(0.0211) (0.0285) (0.0273) (0.0260) (0.0268) (0.0222) (0.0237) (0.0214) (0.0203) (0.0221)N 4842 4958 5307 5452 6010 6967 7048 9897 9886 10751

Panel B1: Mortgage Conditional on No Pension65-79 -0.0784*** -0.0325 -0.139*** -0.170*** -0.0760*** -0.0875*** -0.0339 -0.0917*** -0.0125 0.0296**

(0.0225) (0.0203) (0.0215) (0.0203) (0.0222) (0.0204) (0.0219) (0.0157) (0.0160) (0.0150)80-100 -0.185*** -0.120*** -0.248*** -0.219*** -0.238*** -0.201*** -0.272*** -0.143*** -0.111*** -0.0988***

(0.0232) (0.0201) (0.0219) (0.0242) (0.0217) (0.0235) (0.0224) (0.0223) (0.0224) (0.0196)N 2950 3923 4393 4606 4397 4353 4610 6492 6285 6818

Panel A2: Mortgage Conditional on Pension, Homeowners65-79 -0.0895*** -0.259*** -0.291*** -0.215*** -0.155*** -0.181*** -0.181*** -0.136*** -0.157*** -0.123***

(0.0239) (0.0238) (0.0226) (0.0231) (0.0227) (0.0205) (0.0203) (0.0148) (0.0150) (0.0153)80-100 -0.262*** -0.303*** -0.393*** -0.318*** -0.257*** -0.442*** -0.388*** -0.320*** -0.307*** -0.254***

(0.0229) (0.0333) (0.0372) (0.0312) (0.0321) (0.0244) (0.0276) (0.0230) (0.0233) (0.0234)N 4316 4394 4758 4798 5318 6343 6278 8579 8675 9405

Panel B2: Mortgage Conditional on No Pension, Homeowners65-79 -0.126*** -0.0567** -0.183*** -0.234*** -0.159*** -0.173*** -0.122*** -0.172*** -0.158*** -0.0571**

(0.0335) (0.0276) (0.0283) (0.0272) (0.0278) (0.0280) (0.0279) (0.0203) (0.0219) (0.0222)80-100 -0.268*** -0.165*** -0.309*** -0.292*** -0.302*** -0.315*** -0.387*** -0.230*** -0.286*** -0.220***

(0.0340) (0.0285) (0.0288) (0.0320) (0.0307) (0.0336) (0.0309) (0.0298) (0.0293) (0.0265)N 2271 3127 3717 3607 3400 3437 3765 4760 4436 4752

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 7. All models include controls for race, marital status,

and high school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights are

used in all specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

46

Table 11: SCF: Education Heterogeniety in Mortgage Rate by Age Cohort over Time

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)1989 1992 1995 1998 2001 2004 2007 2010 2013 2016

Panel A1: Mortgage Conditional on HS Education or Less65-79 -0.0902*** -0.119*** -0.211*** -0.131*** -0.0851*** -0.0728*** -0.0938*** -0.119*** -0.0251 -0.0236**

(0.0183) (0.0190) (0.0190) (0.0188) (0.0209) (0.0196) (0.0203) (0.0154) (0.0155) (0.0114)80-100 -0.201*** -0.173*** -0.316*** -0.200*** -0.203*** -0.238*** -0.322*** -0.171*** -0.0927*** -0.142***

(0.0175) (0.0205) (0.0176) (0.0215) (0.0220) (0.0200) (0.0215) (0.0210) (0.0202) (0.0162)N 3994 3793 3998 3883 3779 3764 3918 5987 5790 13649

Panel B1: Mortgage Conditional on More than HS Education65-79 -0.165*** -0.218*** -0.197*** -0.198*** -0.183*** -0.176*** -0.0976*** -0.0470*** -0.0672*** -0.00262

(0.0282) (0.0254) (0.0236) (0.0250) (0.0226) (0.0204) (0.0214) (0.0151) (0.0149) (0.0273)80-100 -0.301*** -0.386*** -0.302*** -0.295*** -0.327*** -0.331*** -0.293*** -0.250*** -0.275*** -0.129***

(0.0387) (0.0309) (0.0354) (0.0323) (0.0270) (0.0295) (0.0271) (0.0235) (0.0209) (0.0429)N 3798 5088 5702 6175 6628 7556 7740 10402 10381 3920

Panel A2: Mortgage Conditional on HS Education or Less, Homeowners65-79 -0.105*** -0.178*** -0.264*** -0.201*** -0.150*** -0.133*** -0.189*** -0.232*** -0.190*** -0.105***

(0.0246) (0.0243) (0.0230) (0.0245) (0.0250) (0.0258) (0.0245) (0.0184) (0.0201) (0.0142)80-100 -0.252*** -0.231*** -0.410*** -0.297*** -0.253*** -0.345*** -0.433*** -0.281*** -0.258*** -0.263***

(0.0231) (0.0278) (0.0218) (0.0293) (0.0285) (0.0278) (0.0276) (0.0253) (0.0262) (0.0191)N 3113 2909 3219 2840 2815 2851 3014 4366 4096 10537

Panel B2: Mortgage Conditional on More than HS Education, Homeowners65-79 -0.141*** -0.281*** -0.235*** -0.247*** -0.165*** -0.216*** -0.135*** -0.0699*** -0.141*** -0.0656**

(0.0329) (0.0282) (0.0258) (0.0261) (0.0250) (0.0219) (0.0221) (0.0155) (0.0157) (0.0274)80-100 -0.324*** -0.445*** -0.300*** -0.345*** -0.344*** -0.401*** -0.326*** -0.300*** -0.359*** -0.189***

(0.0448) (0.0379) (0.0454) (0.0365) (0.0317) (0.0326) (0.0336) (0.0269) (0.0256) (0.0477)N 3474 4612 5256 5565 5903 6929 7029 8973 9015 3620

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 7. All models include controls for race, marital

status, and high school education indicators, as well as logged income, number of children, and net worth quartiles. In the education

split, we do not control for education. Household weights are used in all specifications. Excluded age group is 50-64. * p<0.10, **

p<0.05, *** p<0.01

47

Table 12: HRS: Wealth Heterogeneity in Mortgage Rate by Age Cohort over Time

(1) (2) (3) (4) (5) (6) (7) (8) (9)1998 2000 2002 2004 2006 2008 2010 2012 2014

Panel A1: Mortgage Conditional on Above Median Wealth65-79 -0.224*** -0.205*** -0.224*** -0.239*** -0.177*** -0.198*** -0.234*** -0.243*** -0.200***

(0.021) (0.024) (0.025) (0.025) (0.027) (0.029) (0.025) (0.026) (0.027)80-100 -0.284*** -0.260*** -0.279*** -0.336*** -0.305*** -0.323*** -0.362*** -0.379*** -0.354***

(0.023) (0.025) (0.027) (0.027) (0.030) (0.031) (0.026) (0.027) (0.028)N 3972 3221 2882 3089 2803 2593 3497 3216 3006

Panel B1: Mortgage Conditional on Below Median Wealth65-79 -0.107*** -0.101*** -0.088*** -0.114*** -0.084*** -0.097*** -0.066*** -0.005 -0.007

(0.016) (0.019) (0.019) (0.019) (0.020) (0.022) (0.021) (0.022) (0.023)80-100 -0.173*** -0.187*** -0.183*** -0.233*** -0.211*** -0.220*** -0.196*** -0.120*** -0.108***

(0.017) (0.019) (0.020) (0.020) (0.021) (0.024) (0.021) (0.023) (0.024)N 5028 4124 3736 4055 3639 3304 4671 4275 3868

Panel A2: Mortgage Conditional on Above Median Wealth, Homeowners65-79 -0.237*** -0.213*** -0.231*** -0.245*** -0.176*** -0.193*** -0.248*** -0.251*** -0.214***

(0.022) (0.024) (0.025) (0.025) (0.028) (0.029) (0.026) (0.027) (0.028)80-100 -0.299*** -0.265*** -0.283*** -0.342*** -0.308*** -0.314*** -0.366*** -0.392*** -0.368***

(0.024) (0.027) (0.028) (0.028) (0.031) (0.033) (0.028) (0.028) (0.029)N 3712 2994 2684 2895 2637 2459 3270 3007 2818

Panel B2: Mortgage Conditional on Below Median Wealth, Homeowners65-79 -0.191*** -0.148*** -0.125*** -0.169*** -0.153*** -0.148*** -0.274*** -0.211*** -0.239***

(0.026) (0.030) (0.029) (0.029) (0.031) (0.033) (0.030) (0.031) (0.035)80-100 -0.236*** -0.200*** -0.182*** -0.300*** -0.324*** -0.315*** -0.484*** -0.368*** -0.352***

(0.032) (0.035) (0.037) (0.035) (0.039) (0.043) (0.035) (0.039) (0.042)N 2750 2231 2052 2140 1860 1727 1984 1807 1620

Notes: Health and Retirement Study. This table includes the estimates for Figure 7. All models include state fixed effects, as well as

controls include number of children, logged income, greater than high school education, race, and marital status. Excluded age group

is 50-64. Household weights are used in all specifications. * p<0.10, ** p<0.05, *** p<0.01

48

Table 13: Mortgage Interest Tax Treatment and Mortgage Rate: Household Fixed Effects

(1) (2) (3) (4)Base + MID Base + MID

Age 65-79 -0.017*** -0.018** -0.032*** -0.037***(0.005) (0.007) (0.007) (0.010)

Age 80-100 0.012 0.007 0.024** 0.006(0.009) (0.013) (0.012) (0.018)

65-79 X MID -0.002 -0.001(0.002) (0.003)

80-100 X MID 0.003 0.008*(0.003) (0.005)

MID -0.004 -0.013**(0.003) (0.005)

Always Own X XMSA FE X X X XYear FE X X X XControls X X X XObservations 71,327 71,327 41,063 41,063* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Each observation is a respondent-year. All models include MSA andyear fixed effects and control for number of children, logged income, top state marginal income tax rate,urban area, greater than high school education, race, marital status, bequest motive, net worth quartiles,and subjective health measure. MID is the average marginal mortgage interest tax rate by state and yearfrom the NBER.

49

Table 14: Mortgage Rate by State Unemployment and MSA Rent to Price Ratio (RTP):Household Fixed Effects

(1) (2) (3) (4)Base +UE, RTP Base +UE, RTP

Age 65-79 -0.004 -0.019 -0.009 -0.028(0.006) (0.028) (0.008) (0.039)

Age 80-100 0.015 0.069 0.039** 0.105*(0.011) (0.042) (0.016) (0.063)

65-79 X Unemployment 0.005** 0.007**(0.002) (0.003)

80-100 X Unemployment 0.010*** 0.016***(0.004) (0.005)

65-79 X RTP -0.002 -0.004(0.005) (0.007)

80-100 X RTP -0.025*** -0.036***(0.008) (0.012)

Unemployment -0.008*** -0.012***(0.002) (0.002)

RTP 0.023** 0.045***(0.010) (0.014)

Always Own X XMSA FE X X X XYear FE X X X XControls X X X XObservations 42,927 42,927 25,009 25,009* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Each observation is a respondent-year. All models include MSA andyear fixed effects and control for number of children, logged income, urban area, greater than high schooleducation, race, marital status, bequest motive, net worth quartiles, and subjective health measure.Unemployment rate is the state unemployment rate that year. RTP is rent to price ratio as estimated byCampbell et al. (2009).

50

Table 15: 60 Days or More Late on Account by Age Cohort over Time

1995 1998 2001 2004 2007 2010 2013 2016Panel A: Full Sample65-79 -0.0148*** -0.0519*** -0.0469*** -0.0248*** -0.0237*** -0.0486*** -0.0335*** -0.0158***

(0.00503) (0.00518) (0.00424) (0.00507) (0.00476) (0.00435) (0.00394) (0.00376)80-100 -0.0152** -0.0583*** -0.0540*** -0.0409*** -0.0447*** -0.0693*** -0.0528*** -0.0476***

(0.00762) (0.00515) (0.00380) (0.00521) (0.00358) (0.00442) (0.00370) (0.00248)N 10140 10329 10661 11736 12040 17050 16745 18210

Panel B: Homeowners Only65-79 -0.00821* -0.0522*** -0.0352*** -0.0133** -0.00951** -0.0476*** -0.0236*** -0.00999**

(0.00498) (0.00593) (0.00425) (0.00541) (0.00485) (0.00474) (0.00416) (0.00413)80-100 -0.00996 -0.0615*** -0.0397*** -0.0266*** -0.0316*** -0.0651*** -0.0431*** -0.0387***

(0.00790) (0.00540) (0.00383) (0.00602) (0.00315) (0.00468) (0.00329) (0.00274)N 8855 8591 8917 10084 10350 13835 13540 14685

Panel C: Homeowners Only, with Controls65-79 0.00813 -0.0516*** -0.0293*** -0.00843 -0.0155*** -0.0461*** -0.0122** -0.00809*

(0.00604) (0.00734) (0.00612) (0.00661) (0.00596) (0.00581) (0.00494) (0.00473)80-100 0.0133 -0.0550*** -0.0256*** -0.0189** -0.0388*** -0.0601*** -0.0225*** -0.0353***

(0.00806) (0.00795) (0.00636) (0.00852) (0.00598) (0.00674) (0.00503) (0.00426)N 8475 8405 8718 9780 10043 13339 13111 14157

Notes: Survey of Consumer Finances. This table includes the estimates for Figure 10. All models include controls for race, marital

status, and high school education indicators, as well as logged income, number of children, and net worth quartiles. Household weights

are used in all specifications. Excluded age group is 50-64. * p<0.10, ** p<0.05, *** p<0.01

51

Table 16: Mortgage Rate by Mortgage Interest Tax Treatment: Initial State of Residence

(1) (2) (3) (4)Base + MID Base + MID

Age 65-79 -0.148*** -0.147*** -0.017*** -0.020***(0.004) (0.007) (0.005) (0.007)

Age 80-100 -0.250*** -0.247*** 0.012 -0.004(0.005) (0.009) (0.009) (0.013)

65-79 X MID -0.006*** -0.002(0.002) (0.002)

80-100 X MID 0.002 0.004(0.003) (0.003)

MID -0.000 -0.003(0.002) (0.004)

HH FE X XMSA FE X X X XYear FE X X X XControls X X X XN 71,327 71,327 71,327 71,327* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Linear probability models with robust standard errors clustered atthe state-level in parentheses. Each observation is a respondent-year. All models include MSA and yearfixed effects and control for number of children, logged income, top state marginal income tax rate, urbanarea, greater than high school education, race, marital status, bequest motive,net worth quartiles, andsubjective health measure. MID is the average marginal mortgage interest tax rate by state and year fromthe NBER based on initial state of residence.

52

Table 17: Mortgage Rate by State Unemployment and MSA Rent to Price Ratio (RTP):Initial State of Residence

(1) (2) (3) (4)Base + RTP, UE Base + RTP, UE

Age 65-79 -0.166*** -0.194*** -0.004 -0.038(0.005) (0.028) (0.006) (0.027)

Age 80-100 -0.261*** -0.367*** 0.015 0.050(0.007) (0.036) (0.011) (0.043)

65-79 X Unemployment 0.009*** 0.007***(0.002) (0.002)

80-100 X Unemployment 0.015*** 0.011***(0.003) (0.004)

65-79 X RTP -0.005 0.001(0.005) (0.005)

80-100 X RTP 0.007 -0.022***(0.007) (0.008)

Unemployment -0.007*** -0.009***(0.002) (0.002)

RTP 0.031*** 0.019*(0.009) (0.010)

HH FE X XMSA FE X X X XYear FE X X X XControls X X X XN 42,927 42,928 42,927 42,928* p<0.10, ** p<0.05, *** p<0.01

Notes: Health and Retirement Study. Linear probability models with robust standard errors clustered atthe state-level in parentheses. Each observation is a respondent-year. All models include MSA and yearfixed effects and control for number of children, logged income, urban area, greater than high schooleducation, race, marital status, bequest motive, net worth quartiles, and subjective health measure.Unemployment rate is the state unemployment rate that year. RTP is rent to price ratio as estimated byCampbell et al. (2009).

53